Updated the Capturing procotol in the Recording package target to pass captured audio as a file URL instead of data.

This commit is contained in:
2026-07-04 15:06:55 +02:00
parent 2fe240de1a
commit 6a0e7698a5
10 changed files with 116 additions and 119 deletions
@@ -19,7 +19,7 @@ public protocol Capturing: Sendable {
/// Stops the recording. /// Stops the recording.
/// ///
/// - Returns: The audio captured since the recording started. /// - Returns: The location of the audio file captured since the recording started.
func stop() async throws -> Data func stop() async throws -> URL
} }
@@ -8,14 +8,14 @@ public protocol Transcribing: Sendable {
// MARK: Methods // MARK: Methods
/// Transcribes the given recorded audio into text, letting the service be called directly as a function. /// Transcribes the given recorded audio file into text, letting the service be called directly as a function.
/// ///
/// - Parameters: /// - Parameters:
/// - audio: The recorded audio to transcribe. /// - audio: The location of the recorded audio file to transcribe.
/// - locale: The locale of the spoken language to transcribe. /// - locale: The locale of the spoken language to transcribe.
/// - Returns: The transcription of the recorded audio. /// - Returns: The transcription of the recorded audio.
func callAsFunction( func callAsFunction(
_ audio: Data, _ audio: URL,
locale: Locale locale: Locale
) async throws -> Transcription ) async throws -> Transcription
@@ -2,7 +2,7 @@ import AVFoundation
/// The recording service that captures real audio from the device's microphone. /// The recording service that captures real audio from the device's microphone.
/// ///
/// The service records into a temporary `.m4a` file through an `AVAudioRecorder`, and returns the file's contents when the recording stops. /// The service records into a temporary `.m4a` file through an `AVAudioRecorder`, and returns the file's location when the recording stops.
/// The user's permission to record is requested before a recording starts and, on the platforms that require it, the shared audio session is /// The user's permission to record is requested before a recording starts and, on the platforms that require it, the shared audio session is
/// configured for recording while the capture is in progress. /// configured for recording while the capture is in progress.
@MainActor @MainActor
@@ -73,12 +73,11 @@ public final class AudioCapturing: Capturing {
} }
} }
/// Stops the recording, deleting the temporary audio file after its contents are read. /// Stops the recording, handing the temporary audio file over to the caller, which becomes responsible for deleting it.
/// ///
/// - Returns: The audio captured since the recording started. /// - Returns: The location of the audio file captured since the recording started.
/// - Throws: ``AudioCapturingError/noOngoingRecording`` when no recording is in progress, or any error thrown /// - Throws: ``AudioCapturingError/noOngoingRecording`` when no recording is in progress.
/// while reading the recorded audio file. public func stop() async throws -> URL {
public func stop() async throws -> Data {
guard let recorder else { guard let recorder else {
throw AudioCapturingError.noOngoingRecording throw AudioCapturingError.noOngoingRecording
} }
@@ -91,11 +90,7 @@ public final class AudioCapturing: Capturing {
try? AVAudioSession.sharedInstance().setActive(false) try? AVAudioSession.sharedInstance().setActive(false)
#endif #endif
defer { return Constant.File.url
try? FileManager.default.removeItem(at: Constant.File.url)
}
return try Data(contentsOf: Constant.File.url)
} }
} }
@@ -3,10 +3,10 @@ import Speech
/// The transcribing service that transcribes recorded audio into text on device. /// The transcribing service that transcribes recorded audio into text on device.
/// ///
/// The service writes the recorded audio into a temporary `.m4a` file and runs it through a `SpeechAnalyzer` with a `SpeechTranscriber` /// The service runs the given recorded audio file through a `SpeechAnalyzer` with a `SpeechTranscriber` module, joining the finalized
/// module, joining the finalized results into the returned transcription. The transcription happens in the locale given at each call or rather in the /// results into the returned transcription. The transcription happens in the locale given at each call or rather in the closest equivalent
/// closest equivalent the transcriber supports. The speech model assets for that locale are downloaded and installed on first use; every /// the transcriber supports. The speech model assets for that locale are downloaded and installed on first use; every transcription after
/// transcription after that happens entirely offline. /// that happens entirely offline.
public struct AudioTranscribing: Transcribing { public struct AudioTranscribing: Transcribing {
// MARK: Initializers // MARK: Initializers
@@ -16,22 +16,20 @@ public struct AudioTranscribing: Transcribing {
// MARK: Methods // MARK: Methods
/// Transcribes the given recorded audio into text, deleting the temporary audio file when the transcription finishes. /// Transcribes the given recorded audio file into text, deleting the file when the transcription finishes.
/// ///
/// - Parameters: /// - Parameters:
/// - audio: The recorded audio to transcribe. /// - audio: The location of the recorded audio file to transcribe.
/// - locale: The locale of the spoken language to transcribe. /// - locale: The locale of the spoken language to transcribe.
/// - Returns: The transcription of the recorded audio. /// - Returns: The transcription of the recorded audio.
/// - Throws: ``AudioTranscribingError/localeNotSupported`` when the transcriber supports no equivalent of the given locale, /// - Throws: ``AudioTranscribingError/localeNotSupported`` when the transcriber supports no equivalent of the given locale,
/// or any error thrown while installing the speech model assets, reading the audio file, or analyzing its contents. /// or any error thrown while installing the speech model assets, reading the audio file, or analyzing its contents.
public func callAsFunction( public func callAsFunction(
_ audio: Data, _ audio: URL,
locale: Locale locale: Locale
) async throws -> Transcription { ) async throws -> Transcription {
try audio.write(to: Constant.File.url)
defer { defer {
try? FileManager.default.removeItem(at: Constant.File.url) try? FileManager.default.removeItem(at: audio)
} }
guard let locale = await SpeechTranscriber.supportedLocale( guard let locale = await SpeechTranscriber.supportedLocale(
@@ -55,7 +53,7 @@ public struct AudioTranscribing: Transcribing {
text += String(result.text.characters) text += String(result.text.characters)
} }
let file = try AVAudioFile(forReading: Constant.File.url) let file = try AVAudioFile(forReading: audio)
if let lastSampleTime = try await analyzer.analyzeSequence(from: file) { if let lastSampleTime = try await analyzer.analyzeSequence(from: file) {
try await analyzer.finalizeAndFinish(through: lastSampleTime) try await analyzer.finalizeAndFinish(through: lastSampleTime)
@@ -75,14 +73,3 @@ public enum AudioTranscribingError: Error {
/// The transcriber supports no equivalent of the locale the transcription was requested with. /// The transcriber supports no equivalent of the locale the transcription was requested with.
case localeNotSupported case localeNotSupported
} }
// MARK: - Constants
/// The constant values used across the audio transcribing service.
private enum Constant {
/// The file constants.
enum File {
/// The location of the temporary file the recorded audio is written into for the analysis.
static let url = FileManager.default.temporaryDirectory.appending(path: "transcription.m4a")
}
}
@@ -19,9 +19,20 @@ struct DummyCapturing: Capturing {
/// Simulates the stop of a recording. /// Simulates the stop of a recording.
/// ///
/// - Returns: An empty audio payload. /// - Returns: The location of a dummy audio file, which is never actually created.
func stop() async throws -> Data { func stop() async throws -> URL {
.init() FileManager.default.temporaryDirectory.appending(path: Constant.File.name)
} }
} }
// MARK: - Constants
/// The constant values used across the recording services.
private enum Constant {
/// The file constants.
enum File {
/// The name of the dummy audio file.
static let name = "dummy.m4a"
}
}
@@ -11,11 +11,11 @@ struct DummyTranscribing: Transcribing {
/// Simulates the transcription of the given recorded audio with a two-second delay. /// Simulates the transcription of the given recorded audio with a two-second delay.
/// ///
/// - Parameters: /// - Parameters:
/// - audio: The recorded audio to transcribe. /// - audio: The location of the recorded audio file to transcribe.
/// - locale: The locale of the spoken language to transcribe. /// - locale: The locale of the spoken language to transcribe.
/// - Returns: A dummy transcription. /// - Returns: A dummy transcription.
func callAsFunction( func callAsFunction(
_ audio: Data, _ audio: URL,
locale: Locale locale: Locale
) async throws -> Transcription { ) async throws -> Transcription {
try await Task.sleep(for: Constant.Delay.transcribing) try await Task.sleep(for: Constant.Delay.transcribing)
@@ -0,0 +1,48 @@
import Foundation
import Recording
/// A recording service that records its invocations in call order, each one taking a configurable amount of time, and can be
/// configured to fail.
@MainActor
final class CapturingMock: Capturing {
/// The duration every method takes before returning, simulating slow capture work.
var delay: Duration = .zero
/// The error the service throws from every method, or `nil` when it should succeed.
var error: Error?
/// The names of the methods called on the service, in call order.
private(set) var calls: [String] = []
func start() async throws {
try await called("start")
}
func pause() async throws {
try await called("pause")
}
func resume() async throws {
try await called("resume")
}
func stop() async throws -> URL {
try await called("stop")
return FileManager.default.temporaryDirectory.appending(path: "mock.m4a")
}
private func called(
_ name: String
) async throws {
calls.append(name)
try await Task.sleep(for: delay)
if let error {
throw error
}
}
}
@@ -0,0 +1,2 @@
/// An error to configure the service mocks with.
struct ErrorMock: Error {}
@@ -0,0 +1,30 @@
import Foundation
import Recording
/// A transcribing service with a configurable delay, output, and failure.
@MainActor
final class TranscribingMock: Transcribing {
/// The duration of the simulated transcription work.
var delay: Duration = .seconds(0.2)
/// The error the service throws, or `nil` when it should succeed.
var error: Error?
/// The transcription the service returns.
var transcription = "This is a mocked transcription."
func callAsFunction(
_ audio: URL,
locale: Locale
) async throws -> Transcription {
try await Task.sleep(for: delay)
if let error {
throw error
}
return .init(text: transcription)
}
}
@@ -436,79 +436,3 @@ private extension RecordingView.Model {
} }
} }
// MARK: - Mocks
/// A recording service that records its invocations in call order, each one taking a configurable amount of time, and can be
/// configured to fail.
@MainActor
private final class CapturingMock: Capturing {
/// The duration every method takes before returning, simulating slow capture work.
var delay: Duration = .zero
/// The error the service throws from every method, or `nil` when it should succeed.
var error: Error?
/// The names of the methods called on the service, in call order.
private(set) var calls: [String] = []
func start() async throws {
try await called("start")
}
func pause() async throws {
try await called("pause")
}
func resume() async throws {
try await called("resume")
}
func stop() async throws -> Data {
try await called("stop")
return .init()
}
private func called(
_ name: String
) async throws {
calls.append(name)
try await Task.sleep(for: delay)
if let error {
throw error
}
}
}
/// A transcribing service with a configurable delay, output, and failure.
@MainActor
private final class TranscribingMock: Transcribing {
/// The duration of the simulated transcription work.
var delay: Duration = .seconds(0.2)
/// The error the service throws, or `nil` when it should succeed.
var error: Error?
/// The transcription the service returns.
var transcription = "This is a mocked transcription."
func callAsFunction(_ audio: Data, locale: Locale) async throws -> Transcription {
try await Task.sleep(for: delay)
if let error {
throw error
}
return .init(text: transcription)
}
}
/// An error to configure the service mocks with.
private struct ErrorMock: Error {}